Execution Layer Meeting 163

[timbeiko]

Meeting Info

• June 8, 2023, 14:00-15:30 UTC
• Youtube Stream: https://youtube.com/live/LqaR-kdnOoU
• Zoom: shared on #allcoredevs Discord channel shortly before the call
• Ethereum Protocol Calls Calendar subscription

Agenda

• Cancun Progress
  • 4844 blob limit
  • devnet-6
• Cancun scope planning
  • EIP-4788 design
• Engine API: employ one method one structure approach for V3 execution-apis#418
• Honest validator duties (context)
• Holesky Testnet Launch Coordination Call #0 #803

[ralexstokes]

I'd like to raise one final question about EIP-4788: the current design has unbounded storage growth; a simple fix is to swap how the precompile keeps the block root history and bound the insertions by time. however, if we maintain the invariant that the EL does not know about SECONDS_PER_SLOT then swapping how the precompile is indexed would mean SECONDS_PER_SLOT wasted writes that in practice clients will likely optimize anyway with the CL constant.

I'd like to make a decision on this: how bad is the state growth (at say 80MB per year) and does it motivate the other design option?

From there, I'd like to move EIP-4788 to "included" status for Cancun -- we are trying to put together the Deneb spec on the CL and this cross-layer EIP is a small wrinkle while its status is up in the air.

[mkalinin]

I'd like to make a quick update on Engine API

• Engine API: employ one method one structure approach for V3 execution-apis#418

[dankrad]

I already presented this on ACDC last week so like to keep this short. I want to make the case to increase the 4844 limits to 6 blobs max and 3 blobs target.

Data from my experiments sending blocks with extra data between 128 kB and 1 MB on mainnet: https://notes.ethereum.org/OZJ90ZlcQtyrtZa6INYixQ?view

Network was stable in all conditions, one block was reorged due to being late in one of the 1 MB experiments. (This block was just past the 4s deadline on most sentry nodes; note that we often see blocks even at 3.5s even without large blocks being involved)

[tbenr]

I remember we discussed in the past about what should be the honest validator behaviour wrt data availability and duty production. When a fresh node joins the network from a checkpoint sync, assuming EL is already in sync, can it start attesting as soon as the head is reached or should it have to backfill and validate the entire data availability window before?
Shouldn't we be clear about this in honest validator specs?

This is more CL discussion btw.. my bad.

[timbeiko]

Added everything here! @tbenr I've put yours last, so if we have time we can go over it, otherwise, we can discuss on ACDC next week.

[barnabasbusa]

A bit off topic, but Holesky first discussion is happening next week thursday: #803

It would be great if we could get at least one person to show up from each client teams, and we would like to see every large node operators to participate also.

[timbeiko]

Closed in favor of #808